trinhkhng/slerp_Merged_Qwen2-0.5B_0.3

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 6, 2026Architecture:Transformer Featherless Exclusive Cold

trinhkhng/slerp_Merged_Qwen2-0.5B_0.3 is a 0.5 billion parameter language model merged from two Qwen2-0.5B variants using the SLERP method with a parameter of 0.3. This model combines the base Qwen2-0.5B with a debiased version, aiming to integrate their respective characteristics. With a context length of 32768 tokens, it is suitable for tasks requiring moderate context understanding and potentially reduced bias.

Loading preview...

Model Overview

trinhkhng/slerp_Merged_Qwen2-0.5B_0.3 is a 0.5 billion parameter language model created by trinhkhng through a merge operation. This model leverages the SLERP (Spherical Linear Interpolation) merge method, combining two distinct versions of the Qwen2-0.5B architecture.

Merge Details

The model was constructed by merging a base /kaggle/working/Qwen2-0.5B model with a /kaggle/working/debias_Qwen2-0.5B variant. The SLERP method was applied with a specific interpolation parameter t: 0.3, indicating a weighted combination of the two source models. This approach aims to blend the characteristics of the base model with those of the debiased version.

Key Characteristics

  • Architecture: Based on the Qwen2-0.5B family.
  • Parameter Count: 0.5 billion parameters, offering a compact size for efficient deployment.
  • Context Length: Supports a substantial context window of 32768 tokens, enabling processing of longer inputs.
  • Merge Method: Utilizes the SLERP technique, known for smoothly interpolating between model weights.

Potential Use Cases

This merged model is suitable for applications where a balance between performance, efficiency, and potentially reduced bias (due to the inclusion of a debiased source model) is desired. Its moderate size makes it a candidate for tasks that do not require the scale of larger models but benefit from a robust base architecture and extended context.